23 samples and 20394 genes. This means that we
were left with high-quality data for 23 samples, each of which had
expression measurements for 20394 genes.logFC, log fold change . This ranked list has been
stored in the file ./A3/pSS.rnk for further investigation
in this assignment.library(ggplot2) #ploting
library(stringr) #string processing#' Plot out dot plot for enrichment analysis results
#'
#' @param data enrichment result
#'
#' @return dot plot
dotplot <- function(data){
plot <- ggplot(data) +
geom_point(aes(
x = es,
color = FDR,
y = term_name,
size = size)) +
theme(axis.title.x = element_text(size = 8),
axis.title.y = element_text(size = 8)) +
scale_color_gradient(low = "red", high = "blue") +
labs(x = "Enrichment Score(ES)",
color = "P-value",
size = "Geneset Size",
y = NULL)
return(plot)
}4.3.2, a well-established and
widely-used computational method to identify and interpret the
functional context of differentially expressed genes. The Bader Lab gene
set, known for its comprehensive and up-to-date collection of gene sets,
will be utilized in its most recent release dated
March 2, 2023, ensuring that our analysis is based on the
latest available knowledge.pos_path <- "./A3/gsea_report_for_pos_pSS.tsv"
neg_path <- "./A3/gsea_report_for_neg_pSS.tsv"
pos_enrich <- read.table(file = pos_path, sep = '\t', header = TRUE, fill = TRUE)
neg_enrich <- read.table(file = neg_path, sep = '\t', header = TRUE, fill = TRUE)Enrichment in pSS group:
# perpare the input
pos_package <- data.frame(
term_name = str_wrap(sub("\\%.*", "", pos_enrich$NAME), 20), # truncate %
size = pos_enrich$SIZE,
FDR = pos_enrich$FDR.q.val,
es = pos_enrich$ES
)
# plot
dotplot(pos_package[1:10,])
Figure 1. The Dot plot displays the GSEA results for 10 significant hits in enriched genes for pSS group.
Enrichment in non-pSS group:
# prepare the input
neg_package <- data.frame(
term_name = str_wrap(sub("\\%.*", "", neg_enrich$NAME), 20), # truncate %
size = neg_enrich$SIZE,
FDR = neg_enrich$FDR.q.val,
es = neg_enrich$ES
)
# plot
dotplot(neg_package[1:10,])
Figure 2. The Dot plot displays the GSEA results for 10 significant hits in enriched genes for non-pSS group.
We pull the results from ORA analysis from A2:
| Up-regulated Genes | Down-regulated Genes |
|---|---|
| response to other organism | tear secretion |
| response to external biotic stimulus | response to lipopolysaccharide |
| response to biotic stimulus | regulation of epithelial cell proliferation |
| immune system process | organ or tissue specific immune response |
| immune response | neutrophil aggregation |
| antigen processing and presentation of peptide antigen via MHC class I | mucosal immune response. |
| antigen processing and presentation of peptide antigen | intermediate filament organization |
| antigen processing and presentation of endogenous peptide antigen | intermediate filament cvtoskeleton organization |
| antigen processing and presentation of endogenous antigen | innate immune response in mucosa |
| antigen processing and presentation | epithelial cell proliferation |
Table 1. ORA results from A2 for up-regulated genes and down-regulated genes for pSS condition.
By comparing the GSEA results (Figure 1 and Figure 2) and the ORA result (Table 1), the enriched gene sets identified by two distinct analytical methods, demonstrate a strong alignment and consistency in their findings. Both methods reveal significant enrichment in immune response-related gene sets, further substantiating the well-established autoimmune nature of pSS(Luo et al. 2020). On the other hand, for the non-pSS group, the ORA indicates enrichment in cell proliferation, which can be linked to the skin development and protein synthesis gene sets identified in the GSEA results.
However, the comparison is not straightforward. This is primarily due to the fact that the gene sets used in each method are not identical. In the GSEA analysis, we utilized the BaderLab curated gene sets (BaderLab 2016b), which are curated from multiple sources, whereas in the ORA analysis, we only covered Gene Ontology Biological Process (GO:BP) gene sets (Mi et al. 2019). As a result, the conclusions we have drawn are based on a qualitative comparison of gene set names rather than a rigorous quantitative analysis.
Figure 3.
Enrichment map showing pSS vs. non-pSS with threshold FDR at 0.01.
0.01, effectively limiting the
number of gene sets to 745 and edge counts to 5,914, as illustrated in
Figure 3. The edge is constructed
using the default algorithm which the edge value is weighted eqaully by
Jaccard and overlaps. Edge values lower than 0.375 are
filtered out.
Figure 4.
Annotated enrichment map showing the clustered genesets.
We then use AutoAnnotate app(BaderLab 2016a) to construct the annotation after manual adjustment discussed in Journal. MCL clustering algorithm in the app (BaderLab 2016a) is used to show the common labels among nodes. For the labels, we have max word per label leaving default at 3 and minimum word occurrence at 1.
As shown in Figure 4, the analysis of enriched gene sets for pSS samples has identified the regulation of immune response as the most common theme. Additionally, some cell differentiation and proliferation keywords have emerged in our findings, highlighting other potential molecular mechanisms at play in pSS. In contrast, for non-pSS groups, our results show enrichment in gene sets related to protein synthesis and cell structure.
Figure 5. Collapsed theme network reflecting the enriched gene sets for pSS groups.
As shown in Figure 5, we used AutoAnnotate app(BaderLab 2016a) to construct a summary network for the annotated graph. The major themes are consistent with the observation before, including regulation of immune response, cell differentiation and proliferation for pSS group. Some protein synthesis and cell structure themes are related to non-pSS groups.
The results fit our proposed model and major findings in the original publication(Luo et al. 2020), e.g. identification of Immuno response. Interestingly, we identified more themes related with skin, cell structure and differentiation than our expectations. Further investigation is needed to determine the relationship between these molecular landscapes to pSS.
Increasing evidence suggests that pSS may be associated with inheritable factors, and multiple publications have identified potential risk factors linked to mutations in transcription factors (Imgenberg-Kreuz et al. 2021). To further explore this relationship, we performed a post-analysis (signature gene set analysis) with the aim of identifying enriched transcription factors associated with our findings.
For this purpose, we used the signature gene set of Transcription Factors from Bader Lab(BaderLab 2016b), utilizing their latest released version from March 2, 2023. To compute the graph, we employed the EnrichmentMap app(BaderLab 2016a) and applied a two-sided Mann-Whitney test to set the threshold at a p-value of 0.05. This allowed us to identify and showcase three selected signature sets as demonstrative examples of the transcription factors potentially implicated in pSS. These transcription factor signature sets are shown in table Table 2:
| Signiture TFs |
|---|
| IRF5 |
| STAT4 |
| ATF4 |
Table 2. Selected transcription factor genesets for post-analysis.
Figure 6. Selected Transcription Factor signature genesets. (A) STAT4 interacts with multiple immune response genesets. (B) IRF5 is related to protein biosynthesis and cell development. (C) ATF4 enriches to infection signalling pathway.
As shown in Figure 6, we closely examined three transcription factor gene sets: STAT4, IRF5, and ATF4. STAT4 is associated with multiple immune response gene sets, while IRF5 is related to protein synthesis and cell development. Notably, both of these transcription factors have been reported as risk factors for pSS (Imgenberg-Kreuz et al. 2021). IRF5, in particular, is a transcription factor that operates downstream of toll-like receptors (TLRs) and type I IFN receptors, promoting the expression of numerous antiviral and pro-inflammatory proteins(Takaoka et al. 2005). Meanwhile, STAT4 is involved in activating the downstream immunomodulatory cytokine secretion(Nordmark et al. 2009). Our findings align with the known roles of these transcription factors in pSS.
However, to the best of our knowledge, ATF4 has not been reported in any articles related to pSS. ATF4 is known to mediate programmed cell death by promoting the expression of genes involved in cellular amino acid metabolic processes, mRNA translation, and the terminal unfolded protein response (Ohoka et al. 2005). This suggests that ATF4 plays a role in stress response. Intriguingly, the gene is primarily located on the X chromosome (Ohoka et al. 2005), and given pSS’s high occurrence in females (16 females : 1 male) (Imgenberg-Kreuz et al. 2021), this may imply a potential role for ATF4 in pSS pathogenesis. Nevertheless, further investigation is needed to confirm these preliminary observations and to fully understand the role of ATF4 in the context
Our enrichment results lend support to the conclusions and mechanisms discussed in the original paper (Luo et al. 2020). The original paper described pSS as an autoimmune disease, and our findings further emphasize the significant enrichment of immune response in the pSS group. Several other pathogenesis features align with the original paper as well, such as the enrichment of the lymphocytic infiltration gene set and the identification of immune-response-related protein synthesis. When comparing our results with the ORA from the previous assignment, we find that the majority of enrichments are consistent between the two methods, with the exception of pathways in non-pSS groups. The ORA suggests that some pathways enriched in non-pSS groups are related to tear secretion, which is not fully captured in our non-thresholded method. One possible explanation for this discrepancy is that we utilized a more comprehensive and curated gene set from Bader Lab (BaderLab 2016b) in our analysis, allowing us to explore the pathogenesis at a higher resolution.
We would like to highlight a novel finding from our post-analysis (here), which identified a novel transcription factor, ATF4, as being associated with enriched gene sets in pSS, even though it has not been reported in previous pSS studies to the best of our knowledge. Transcription factor ATF4 is known to mediate programmed cell death by promoting the expression of genes involved in cellular amino acid metabolic processes, mRNA translation, and the terminal unfolded protein response (Ohoka et al. 2005). This function aligns with the immune response mechanism of pSS. Additionally, ATF4 has multiple gene replications located on the X chromosome. The X chromosome has been proposed as a promising candidate for involvement in pSS pathogenesis, as the study by Imgenberg et al. (Imgenberg-Kreuz et al. 2021) reported a higher risk for females, with a male-to-female ratio of 1:16. This skewed prevalence suggests a potential connection between the X chromosome and the development of pSS. Taking these factors into account, it may suggest that ATF4, with its roles in cellular processes and location on the X chromosome, may have a potential pathogenic role in pSS. Further investigation is necessary to confirm this hypothesis and elucidate the precise mechanisms through which ATF4 dysfunction might contribute to the development of pSS.